1 For a similar opposition between Saussurean ‘abstract objectivism’ and Vosslerian ‘individual

subjectivism’ see Voloshinov 1973.

Pragmatics 12:4.415-446 (2002)

International Pragmatics Association

ORDERLY AFFECT:

THE SYNTACTIC CODING OF PRAGMATICS

IN WELSH EXPRESSIVE CONSTRUCTIONS

H. Paul Manning

Abstract

This paper describes and analyzes a series of paradigmatic oppositions between N’ constructions in the P-

Celtic languages (Welsh, Breton, Cornish) which serve to code expressive pragmatics of adjectives. The paper

considers both paradigmatic and syntagmatic aspects of these constructions, and shows that asymmetric

interaction of constructions in paradigms influences their purely formal syntagmatic interactions. A typology

of expressive categories is built to serve as a framework for comparison between constructions. It is argued

that a view of grammar that includes both formal and functional dimensions (‘the coding view’) also provides

valuable insight in matters of purely formal constructional interaction.

Keywords: Affect, Syntactic coding, Expressives, Welsh.

1. All subjects leak: On expressives and ‘expressivism’

There are few terms in linguistics that are so multivalent as the term ‘expressive’

(Stankiewicz 1964; Irvine 1982). The term belongs as much to the ‘practical’, ‘folk’

vocabulary of the linguist as it does to the ‘theoretical’, ‘expert’ vocabulary. The various

nuances it has seem to encapsulate an entire Western folk theory of creative subjective

‘expression’, what Taylor (1975: Chapter 1, 1985, 1989: chapter 21) calls ‘expressivism’.

Expressivism is a broadly subject-centered view of language associated with post-

Enlightenment thought (as opposed to object-centered referentialist or ‘designativist’

Enlightenment theories of language, embodied, for example, in truth-conditional semantics

(Taylor 1985: 243)1; that, in its most basic form, privileges those features of language that

embody properties of subjects (creativity, agency, affect, play, poeticity, style, variation,

and so on (Taylor 1975: chapter 1, 1985, 1989: chapter 21)). These opposed theories of

language have long since become theories of opposed sets of phenomena within language

(Taylor 1985). The class of linguistic phenomena called ‘expressives’ usually consists of

all those linguistic phenomena that best embody our intuitive notion of ‘expressivism’ in

language, that is, all those heterogenous linguistic features related to subjective expression

. The coherence of this system of classification does not derive from any coherence in the

416 H. Paul Manning

2 This multifactor definition gives rise to a continuum (compare Bat El 2000: 65), though Zwicky

and Pullum stipulate (ibid.: 338) that it is to be interpreted as a binary distinction, they do not explain how

this can be.

linguistic phenomena denoted (Dressler and Barbaresi 1994: 37), but rather derives from

the folk theory of ‘expressivism’ that involves a whole set of assumptions about the

subject-centered properties of language and their formal expression.

As a result, what is ‘expressive’ is all that which is not grammatical, or marginal,

a ‘fuzzy periphery’ (Stankiewicz 1964: 267), so much so that the term no longer denotes

solely the linguistic coding of affective dispositions of subjects, but a ‘nebulous realm’

consisting of all that does not belong to ordinary grammar or denotational language (Irvine

1990: 41, 46n5). In short, all those places where, as Sapir famously put it, ‘grammars

leak’ (Stankiewicz 1964: 241). ‘Expressives’, both as categories of form and categories of

function, are frequently opposed as a separate (or at the very least ‘peripheral’) competence

alongside ordinary ‘core’ grammatical competence, forming a kind of quasi-natural

‘allolinguistics’ that requires different methodologies and has different lessons than

‘ordinary’ linguistic phenomena (Zwicky and Pullum 1987). The heterogeneity of the class

of phenomena that results from this segregation (Dressler and Barbaresi 1994: 37) is often

taken to be typical of the nature of the phenomena itself: A heterogeneous grouping of

peculiar and variable peripheral phenomena united by their unruly rebellion against core

grammar.

As an unruly and quasi-natural outward objectification of subjective states of affect,

expressives (or emotives, since these are identified in Jakobson’s initial definition (1960:

354; Staniewicz 1964) are often treated as being at the same time a functional category

focusing on speaker affect, and a formal category of ill-behaved devices which

characteristically code this affect, the latter of which are almost all marginal or peripheral

to the ‘cognitive’ grammar. Thus, expressives consist of those places where both

‘grammars’ and ‘subjects leak’, where subjective properties are quasi-naturally objectified

in linguistic form. Most definitions of ‘expressive’ (or ‘emotive’) include, then, some

conflation of functional and formal considerations, even where expressives are still felt to

be part of a conventional grammar (explicitly in Stankiewicz 1964: 229). Most influential

has been the recent definition of ‘expressive morphology’ (as opposed to ‘plain

morphology’) of Zwicky and Pullum 1987, in which pragmatic considerations of

‘expressive, playful, poetic or simply ostentatious effects of some kind’ (ibid.: 335-6) are

considered to be as diagnostic of the underlying difference in competence involved as is

peculiar (ibid.: 336) or variable behavior (ibid.:337) with respect to phonology, morphology

and syntax (ibid.: 336-8). Since none of these criteria, as stated, is sufficient or necessary2

, it is possible for a given phenomenon to be expressive in function, but not in form, or vice

versa, or both.

Since expressives are seemingly indifferent to the opposition of linguistic form and

linguistic function, it becomes possible for one to speak of ‘expressive phenomena’ that do

not express affect or any other subjective property (Irvine 1982: 32) Here, expressives

become a purely formal class. Hence, a whole group of signs, called ‘ideophones’, are also

often called expressives (for example Diffloth 1972; also Zwicky and Pullum 1987: 334-5).

Ironically, ideophones and their ilk (for example, language-games, Zwicky and Pullum

1987: 332-4) usually involve self-conscious modifications of the sign form (for example

Orderly affect 417

atypical phonology and reduplication) that draw attention to the sign form itself, so the

iconism that results is more a product of what Jakobson called the poetic function (a point

Zwicky and Pullum (1987: 332) raise, but do not address), that is, a linguistic function

which draws attention to the ‘palpability of signs [and] deepens the fundamental dichotomy

of signs and objects’ (Jakobson 1960: 356) As Zwicky and Pullum restate it ‘the fact that

[expressive formations] use linguistic resources in ways that grammatical rules do not

enables them to stand out, to call attention to themselves - and so to serve their expressive

function’ (Zwicky and Pullum 1987: 338). In fact, Dressler and Barbaresi (1994: 36) draw

attention to this self-conscious poetic manipulation of linguistic form by calling such

morphology ‘metamorphology’ to replace Zwicky and Pullum’s ‘expressive morphology’.

And it is this poeticization of linguistic form that renders much ‘expressive morphology’

and its ilk marginal or peripheral to grammatical competence, and accounts for this

terminological quibble in which ‘expressives’ do not express affect or other subjective

states.

Yet it is clear that what sorts of formal operation count as ‘expressive’ is not given

once and for all, but is relative to what counts as ‘plain’ morphology in any given grammar

(Baldi and Dawar 2000). If this is so, then ‘plain’ and ‘expressive’ alternants are

nevertheless in relations of systematic and meaningful paradigmatic contrast, better

expressed as intimate paradigmatic relations between unmarked and marked forms within

a grammar, than as estranged relations between grammar-abiding and grammarless forms.

This implies that the opposition between marked expressives and unmarked non-

expressives is the same as the broader opposition between ‘marked periphery’ and

‘unmarked core’ in grammar. Expanding a recent series of arguments broadly associated

with Construction Grammar (CG), I wish to argue that as we transcend the often empty and

circular distinction between expressives and non-expressives as not-grammar versus

grammar, we should at the same time transcend the equally empty and circular distinction

between marked periphery and unmarked core as oppositions between types of competence

(Culicover and Jackendoff 1999; Kay and Fillmore 1999; Kay 2002), while retaining the

distinction between marked and unmarked within a single grammatical competence. I wish

to argue that to do this, we must not only admit, with practitioners of CG, that constructions

are meaningful, but that they are meaningful because they oppose each other

paradigmatically. The distinction between ‘marked’ and ‘unmarked’ that haunts the

terminology of CG implies the comparison of constructions in absentia (negatively, by

paradigmatic opposition, by partial difference) as well as in presentia (positively, by

syntagmatic composition, by partial identity, as insisted by Construction Grammarians).

Syntactic constructions must code not only positive relations (semantically statable in terms

of compositionality) between co-present items (syntactic form), but the formal differences

of arrangement between these co-present items often form the stuff from which meaningful

paradigmatic contrast between constructions (paradigmatic form) can be formed (the sense

of stuff and form intended here is that of Humboldt (1988)).

The goal of this paper is to show that the category of ‘expressive’ can be treated as

being a relatively well-behaved grammatical category with relatively well-behaved formal

exponents, and that these exponents can be found within both morphological and syntactic

paradigmatic oppositions. This is not to say that expressives will not, from time to time,

be expressed in a manner that is grammatically non-canonical, but in other respects they are

yet grammatical. In so doing, I hope to create a more usable metalanguage for the

418 H. Paul Manning

3 Data are cited by language (Middle W[elsh], C[ornish], B[reton]). Welsh (W) texts are cited

according to standard abbreviations as used in Evans (1976), Breton (B) texts are cited according to the

standard abbreviations given in Hemon (1975). Some Welsh examples have been culled from various texts

published over the years in the Bulletin of the Board of Celtic Studies (abbreviated as B) and these are cited

by volume, page (and line, where appropriate). Cornish data cited are from Origo Mundi (O), Passio Nostri

Domini (D) and Ressurexio Nostri Domni (R) (Norris 1968 [1859]), as well as Pryce 1972 [1790]. Numbers

following the textual abbreviation refer to the page and line of the standard edition.

4 Since this paper is not aimed at any particular syntactic formalism, I will use a phrase structure

koine (traditional node labels like N’, NP, AP etc.) throughout to indicate that my phrase structure labels refer

to empirical “phenogrammatical” constituency (constituent structure within the order of Saussurean signifiers,

that is, perceptible formal categories) rather than “tectogrammatical” constituency (constituent structure as

a (partial) diagrammatic icon of semantic (or similar universally stipulable but not immediately empirical)

relations, constituent structure with the order of Saussurean signifieds). For the distinction see Dowty 1996.

An adequate coding statement requires that these two orders scrupulously be kept distinct. My other

abbreviations in the glosses are as follows: DIR (direct relative particle), OBL (Oblique relative particle) (see

Manning 1996 for the distribution), PT (progressive aspect marker or predicating particle), AFF (affirmative

clause marker), SG (singular number), PL (plural number).

5 This partitive construction bears superficial resemblance to similarly partitive constructions used

to code similar pragmatics of affect, such as English a bitch of a problem, so nice of a man, just as some of

the adjective word order inversions found in MB and MC have uncertain parallels in French (e.g. Marouzeau

1923). The exclamative pragmatics of this construction are paralleled in syntactic constructions in other

languages (see, for example Elliot 1974; Rosengren 1994; Maynard 1995; Michaelis and Lambrecht 1996).

Considerations of space and relevance prevents me from making explicit comparisons here.

pragmatics of expressives, and at the same time to show that expressive categories are

susceptible to being coded in regular paradigmatic opposition to non-expressive categories

in syntax, as in morphology. My data for this project is drawn from expressive syntactic

constructions in Middle Welsh, Middle Cornish, and Middle Breton.3

2. Expressive constructions in P-Celtic.

In each of the P-Celtic languages (Middle Welsh, Middle Cornish and Middle Breton,

henceforth MW, MC and MB) there exists a non-maximal nominal phrasal construction

(N’) which concatenates an adjective (A) and a noun (N) using a partitive construction,

represented in a conventional phrase structure diagram as (1), where a is identifiable with

preposition meaning ‘of’ in MB and MC.4 In order to distinguish this sort of N’ (all N’s

in which N is attributively modified by A) from all others, I will label it NA henceforth. The

NA in question (exemplified in (1)) has a peculiar syntax compared to canonical N A

constructions (2), so I will call this the ‘non-canonical’ NA construction.5

(1) [NA A [PP a N]]

(1a) tec a wr

fair ? man

‘A fair man’ W.SG.364

Orderly affect 419

(1b) tek a bren

fair of wood

‘Fair wood’ C.O.2441

(1c) ur brBo a dap

a brave of joke

‘A good joke’ B. SFKH.16

This NA form minimal pairs with a denotationally synonymous NA counterpart (composed

of the same lexemes), which I will call the ‘canonical’ NA construction:

(2) [NA N A]

(2a) teyrndynyon tec

kingly-men fair

‘Fair kingly men’ W.KO.126

(2b) an aval tek

the apple fair

‘The fair apple’ C.O.267

(2c) un den brao

a man brave

‘A brave man’ B.BD.943

Within the NA category in Middle period P-Celtic languages we find ‘paradigms of

constructions’ (oppositions between non-canonical and canonical NA constructions), so that

the form the NA construction itself takes conveys some additional differential and specific

coded paradigmatic value beyond the compositional attribution relation between N and A.

These opposed NA constructions are denotationally synonymous, that is, the compositional

relation of ‘attributive modification’ between (intensions of) N and A is semantically

identical in both constructions.

Form Denotational Value

Canonical [NA N A] Modification of N by A

Non-canonical [NA A [PP a N]] Modification of N by A

The further difference coded by this syntactic opposition is pragmatic. The non-canonical

NA specifically and differentially codes some nuance of expressive, and more specifically

exclamative, pragmatic value, which the canonical N A lacks. Moreover, this pragmatic

value applies only to the adjective, the position of which is distinctive formally.

Form Pragmatic Value

Canonical [NA N A] Neutral pragmatics

420 H. Paul Manning

6 Another MW deviation from partitive expressive structure involves cases of omission of the

preposition a entirely, as in examples like da varchawc ‘good knight’ (W.SG.294), as opposed to the expected

‘da a varchawc’. This occurs only after the adjective da ‘good’, which might be explained plausibly by a

contraction (da a N 6 da: N 6 da N). The covert presence of the preposition a is secured by the lenition

(marchawc 6 varchawc) that it triggers.

Non-canonical [NA A [PP a N]] A pragmatically foregrounded

The form of the non-canonical construction is peculiar and has attracted attention in the

past (Fife 1993). A certain amount of attention has been directed to the identity of the

‘head’ (whether A or N) and the resultant categorial status of the phrase (AP or NP) (Fife

1993). Since neither the form nor function of this construction have ever been fully

described, part of the objective of this paper will be to provide a full description to clarify

this matter. The evidence of a fuller description will show that this construction behaves

in every respect as a variety of N’ in its internal and external syntax. In addition, since any

set of A and N lexemes that can participate in the non-canonical construction can also

participate in the canonical one (but not vice versa), the formal opposition between

syntactic constructions here is meaningful in itself in the same way as lexical or

morphological oppositions are.

A second peculiarity of this construction, for MW at least, is that the partitive

preposition used in this construction (a ‘of’) is found in no other construction in MW, the

usual variant being o ‘of’. It is, however, the normal form of the partitive preposition ‘of’

in MC and MB. By later MW, we see this constructional parochialism being eliminated

in favor of the residual form of the partitive preposition, o ‘of, from’ (3ab). 6 For this

reason I will gloss all occurrences of MW a as ‘of’.

(3a) mawr a ovit

great of grief

‘A great grief’W.SG.252

(3b) mawr o ovit

great of grief

‘A great grief’W.SG.228

As the data above shows, each P-Celtic language shows some reflex of the non-canonical

NA construction, though differing in important points of form and distribution. My

argument for this paper is that a thorough description and sustained analysis of this formal

constructional alternation both in terms of distinctive form and coded content will prove

interesting for a view of syntactic arrangements (‘constructions’) as being a kind of ‘form’

entirely analogous to distinctive morphological ‘form’ in terms of their potentiality for

contributing to differential coding of grammatical categorial ‘meaning’.

My approach, the ‘coding view’, which derives from a theory of grammar somewhat

analogous to ‘Construction Grammar’ (such as Zwicky 1987, 1994; Goldberg 1995;

Fillmore 1999; Kay and Fillmore 1999; Kay 2002), though without any attendant cognitive

or psychological commitments, treats categories of differential perceptible form (formal

categories) and categories of differentially coded ‘meaning’ (grammatical categories) as

Orderly affect 421

7 Sapir’s formulation (“language is a system of phonetic symbols for the expression of communicable

thought and feeling”(1949 [1933]: 7)) captures this mutual and reciprocal conditioning of categories of form

(formal categories) and categories of content (grammatical categories) elegantly, opposing “phonetic symbols”

(and not merely meaningless ‘sound’, but meaningful ‘form’) to “communicable thought and feeling” (not

merely inchoate and formless ‘concepts’, but (formally) communicable concepts). In his brilliant phrasing,

each term contains the opposed term as an ‘internal relation’. Form and meaning are brought together in

‘internal relations’ (where each term contains and conditions (forms) its opposite), rather than merely

conjoined in a kind of ‘external relation’. For a more extensive drawing out of the consequences of this

‘configurational’ view see Sapir (1994: 120).

being in a mutually constituting (‘dialectical’) relationship, so that it is impossible, even in

principle, to make arguments about the one without simultaneously making arguments

about the other (Silverstein 1993). This dialectical mutual constitution of categories of

form and function in language is the central interest of the coding view of grammar, which

is thus neither merely ‘formal’ nor ‘functional’ in any of the senses these have been used

in recent debates.7

My description and analysis will therefore have two related goals. I illustrate,

formally how these constructions interact both paradigmatically (in relations of contrast

captured by the idea of ‘paradigms of constructions’ (Silverstein 1993: 326)) as well as

syntagmatically (section 3). I also illustrate, functionally, how the pragmatic value coded

by this construction foregrounds the expressive potentiality latent in certain adjectives, and

develop a metapragmatic vocabulary to capture some important differences between such

constructions (section 4). This typology I apply to Middle Welsh, with a comparison to

cognate syntactic constructions in Middle Cornish and Middle Breton (section 5.). Since

the opposition between ‘normal’ and ‘expressive’ constructions is a paradigmatic

opposition, I argue that instead of allocating the former to the unmarked ‘core’ and the

latter to the marked ‘periphery’ of grammar, we should instead bring the notion of

paradigms of constructions, as well as the notion of categorial markedness based on

paradigmatic contrast and opposition, back into a unitary theory of grammar, wherein there

is no core and periphery (section 6). The opposition between non-expressive and

expressive is not an opposition between what is grammatical and what is not, it is a

grammatical opposition.

3.1. Formal categories: Internal syntax

In previous descriptions, the structure in question has been described as an “headless NP”,

having undergone a process of “decapitation” (Fife 1993). The partitive structure of the

non-canonical phrase is exocentric (there is a conflict between categorial and structural

intuitive definitions of headedness: The most plausible categorial candidate for ‘head’ (N)

is not structurally dominated by the N’ node, but by a PP node), unlike canonical N As,

which are endocentric. I argue instead that the category I have labeled NA is a type of N’

in its internal syntax, but that its external syntax depends on paradigmatic factors and the

referential status of its head N.

In terms of internal syntax, rules of number agreement can be stated uniformly for

all NA constructions, suggesting their categorial identity. Adjectives in normal N A

constructions optionally agree in number with their N heads. This is true whether the

422 H. Paul Manning

8 Fife (1993: 389) claims the opposite for ModW, namely that “it is the initial Adjectives which

determine the...features of the phrase as a whole”, but it is not clear on what data he bases this claim, and the

MW data is at variance with it.

adjective precedes (the adjective hen is unique in this positioning in MW) (4) or follows

the N head (5):

(4a) hen-yon amser-oed (4b) eu hen pechod-eu

Old-PLtimes-PL their old:SG sins-PL

'Old times' W.BD.106.19 'Their old sins' W.BD.83.28

(5a) wyn bychein (5b) kwyss-eu llydan

lambs:PL small:PL furrows-PL wide:SG

'Little lambs' W.B.2.16.2 'Wide furrows' W.B.2.12.22

The fronted adjective in the non-canonical N A shows the same agreement properties as a

normal adjective in a normal NA: The adjective may show plural agreement with the

following N (6a), or it may not (6b). In this construction as in the others, the N is the head

for the purposes of Plural agreement .8

(6a) drud-yon a veirt-yon (6b) meredic a wyr

brave-PL of bards-PL foolish:SG of men-PL

‘Brave bards’ W.M.163.1 ‘Foolish men’ W.KO.763

The position of the adjective relative to the noun within the N A is irrelevant as far as the

distribution of number features is concerned: The rule can only be phrased with regard to

a structurally generic NA in which the relative structural placement of N and A is irrelevant.

In addition, in both Middle Welsh and Middle Breton the N serving as head of the

NA (as a kind of N’) may be modified recursively by other N’ constructions, leading us to

believe that in both the NA construction is a form of N’ headed by N. There are two

possible loci for such recursive N’ modification, working on the assumption that the NA is

a form of N’, either internally or externally to the NA. In Middle Welsh the locus of such

recursion is internal to the NA construction (the N within the apparent PP), while in Middle

Breton the locus of recursion is external to the NA construction. Once this stipulation (on

stipulation see Zwicky 1994) is made, these modifiers in all other respects behave normally.

Hence, numerals precede the N in normal N’ constructions in both languages. In Middle

Welsh, therefore, they appear within the PP of the NA construction, preceding N:

(7) [NA da [

PPa [N’ dwy ynys]]]

good of two island

‘Two good islands’ W.PKM.45.17

Similarly, in MW all other forms of N’ construction (recursive modifications of N) occur

internally to the NA. Adjectives that cannot participate in the non-canonical construction,

for example, appear in their normal position following the N head:

Orderly affect 423

9 What this data shows (following a suggestion by Zwicky 1994) is that seemingly purely formal

asymmetric interactions between constructions on the syntagmatic plane (of the sort considered by Zwicky

1994) may reproduce grammatical categorial (‘functional’) asymmetries on the paradigmatic level.

Grammatical categories are directly invoked in the organization of formal categories into paradigms, but

grammatical categorial markedness is indirectly involved in resolving formal interactions between

constructions syntagmatically. The asymmetries in formal syntagmatic behavior are intuitively the result of

the requirement imposed by coding to preserve in un-neutralized form the differential structure of this

‘marked’ non-canonical NA. The non-canonical NA formally contrasts paradigmatically with a normal NA

by violating certain expectations of homogeneity of phrasal structure within its category type (it is exocentric

while normal NAs are endocentric), and in particular by the extremely unusual positioning of the adjective.

These features allow it to contrast formally with the canonical N A structure and so code a differential

pragmatic value for the adjective. These stipulations of the non-canonical N A are in conflict with the

requirements of more basic AP constructions for phrasal integrity; here the stipulations of the more specific

‘marked’ non-canonical N A structure are sovereign over the ‘unmarked’ AP construction (it is a “firm

construction” in the sense of Zwicky 1994: 621) insofar as adjective positioning is concerned. Following a

suggestion of Zwicky’s (1994), I will suggest that grammatical categorial markedness guides and regiments

(8) [NA mawr [

PPa [N’ gollet anesgor]]]

big of loss incurable

‘A great incurable loss’ W.YCM.150.23

In the same way, N’ internal PP modifiers (30), such as are complements of N and not A,

appear in their normal position (after the head N they modify).

(9)[NA mawr [PP a [N’ gewilyd [

PPy gwbwl o’th genedyl ]]]]

great of shame to all of-your people

‘A great shame to all of your people’ W.SG.272

In MB, by contrast, numerals appear externally to the construction.

(10) [N’ peswar [NA vad [

PPa gezec]]]

four good of horse

‘Four good horses’ B.SBI.ii.16

This is also true of determiners in MB:

(11) [NP

un [NA Fz [

PPa vechér]]]

a easy of trade

‘An easy trade’ B.MG.115

Further evidence comes from discontinuous Adjective phrases produced by the unusual

location of the A required in the non-canonical construction. One might expect the A to

be attended by its phrasal modifiers in this new position. However, with the exception of

the particle mor, which always attends the adjective it modifies (see example 24 below),

phrasal modifiers of A are placed where they would be in a normal adjectival modification

structure, following the N, in MW (12). The requirements of the more specific marked NA

construction take precedence over the more general requirements of integrity of AP, leading

to discontinuities.9

424 H. Paul Manning

purely formal categorial interaction here. Zwicky has wondered if this formal of unyielding ‘firmness’ (or

its corollary ‘softness’, which determines which construction will ‘win’ in any purely formal competition

between constructions) of competing constructions can be motivated by something other than pure stipulation

(ibid.: 621-2). He has also suggested tentatively that “markedness” in at least something approximating the

sense intended here (Zwicky 1994: 620) correlates well with formal asymmetries of interaction between

constructions (so that ‘unmarked’ constructions in grammatical categorial terms are also ‘elsewhere’, ‘default’,

‘basic’, or ‘fall back’ constructions in formal terms, while ‘marked’ constructions are ‘firm’ constructions in

formal terms (ibid.; also Silverstein 1993: 343). I would further suggest that the data here indicates a way that

paradigmatic asymmetries between constructions can be systematically related to syntagmatic asymmetries

of privilege between them, and that asymmetries of the coding of grammatical categories (markedness) are

reproduced in asymmetries of formal categorial interaction (‘firmness’). Paradigmatic asymmetries of

interaction captured by notions of markedness are immediately syntagmatic asymmetries captured by notions

of firmness, and vice versa.

10 Similarly in the following example, the adjective anhebic ‘unexpected’ is construed

discontinuously with its ‘experiencer’ PP (gan y tylwyth ‘(unexpected) by the family’) and the VN

complement (y wneuthur ‘to do it’), both found inside the ‘lower’ N’.

ys [anhebic a gyflafan gan y tylwyth

is unexpected of crime by the familyy wneuthur] a wnaf i yr awr honn

its do DIR I:do I the hour this

‘It is a crime unexpected by the household to be done that I shall do now.’ W.PKM.43.22

(12) [NA cadarn [PP a [N’ ungwri [

NPyi gydymdeithas ]]]]...

strong of man his company

‘A man strong his loyalty (i.e. of strong loyalty)’ W.PKM.7.22

Here, the NP [y gydymdeithas ] ‘his company (loyalty)’ syntactically appears to modify the

head N ungwr ‘man’ directly, where the possessive (y ‘his’) of the former expression refers

to the latter, giving the meaningless translation ‘a mani of hisi loyalty’. Clearly, the phrase

is intended to mean ‘a mani strong (of) his i loyalty’, with the NP [y gydymdeithas] ‘his

company’ part of an AP headed by cadarn ‘strong’, in turn serving as the modifier of

ungwr ‘man’. This would be paraphrasable by a normal NA with a postposed AP modifier

as in (13).10

(13) [NA ungwri [

APcadarn [

NPyi cydymdeithas]]]

man strong his company

‘A man strong his loyalty, i.e. ‘a man of strong loyalty’

In short, the non-canonical N A construction behaves in its internal syntax precisely as a

normal N’ construction. Once a language-specific stipulation (Zwicky 1994) is made as

to precisely where additional N’ constructions will be realized (either internally (MW) or

externally (MB) to the NA construction), other forms of N’ modification proceed normally.

With respect to determiners and phrasal modifiers of the A in this construction, the former

(the particle mor) always attend the A (example 24), while the latter (phrasal modifiers) as

the more general construction, defer to the requirements of the marked non-canonical NA

construction. Since the non-canonical NA construction stipulates only that the A must

Orderly affect 425

precede the N, the requirements of the integrity of AP must be relaxed, so that phrasal

modifiers are placed where they would appear in a normal N’, following N, resulting in

discontinuous APs. Hence, paradigmatic markedness relations between constructions

(canonical versus non-canonical arrangement of A with attendant unmarked versus marked

pragmatic values that these arrangements code) produces asymmetries of syntactic privilege

between co-occurring constructions (thus, more general constructions dealing with, for

example, the disposition of A modifiers with respect to A must defer to constructions which

are paradigmatically marked, for example, the non-canonical A construction (see Zwicky

1994 for a parallel set of suggestions on paradigmatic markedness as a factor in explaining

syntagmatic constructional asymmetries).

3.2. External syntax

A previous analyst (Fife 1993: 384) has suggested that the A is in fact the head of the non-

canonical construction. If this is so, then presumably the construction must be a

‘projection’ of A in its external syntax as well, if we are to cleave to any form of “strictly

categorial determination” (Michaelis and Lambrecht 1996: 225) of external syntax by

internal syntax. The data from external syntax suggests that at least some examples of this

construction (those with referential N heads) are headed by Ns, having the distribution of

N’, that is, a construction that can indifferently serve in some NP and some AP

distributions. On the other hand, those with non-referential N heads (with the ‘dummy

head’ peth ‘thing) have precisely the distributions of A. The two distributions overlap

(hence an ‘underspecified’ categorial identity as ‘nominal’ may be in order), but the specific

syntagmatic categorial identity of the construction in external syntax (as phrasal projection

of N or A) in some sense depends as much on the categorial identity of its paradigmatic

alternant and the referential properties of its head. We cannot, therefore, predict

syntagmatic identity in external syntax from any principle of strictly categorial

determination (internal syntax). Rather, paradigmatic alternation is the determinative factor

resolving external syntactic identity.

Non-canonical NAs may occur in a wide variety of syntactic positions in earlier

MW, but increasingly they seem to be restricted to appearing as cleftee of a cleft of some

sort (also noted for ModW by Fife 1993: 380). This restriction on external distribution is

a deviation from earlier MW, as well as cognate constructions in B and MC, for which my

data on external syntax is scanty. In MW poetry, we find the non-canonical NA serving as

a normal predicate N’ in copular constructions, with an expressed copula (13), or without

(14).

(13) Cop Pred (N’) Subj

bu [NA tru a dynghetven] anghen gywir

was sad of fate need true

‘A sad fate was true need.’ W.CA. 6.136

(14) Pred (N’) Subj

[NA drwc a serch] hwnnw

bad of love that

426 H. Paul Manning

11 One further constructional peculiarity of the external syntax of this construction is that this kind

of cleft is one of the last environments which shows the absolute initial position version of the copula (ys),

which has been replaced in other MW clefts by nothing (q) and in other copular constructions by a post-

positive form of the copula (yw) (Richards 1940). To the extent that it is limited to such constructions it

becomes a further constructional marker of exclamative pragmatics.

‘That (is) bad love.’ W.KO.465

By MW, we find that the non-canonical NA is more or less completely limited to cleft

position (which may be treated as kind of predicate position syntactically). Since virtually

any phrasal constituent type may be clefted in constructions of one sort or another in MW

(Richards 1940; Manning 1996), this does not resolve matters per se.11

This restriction to the pragmatically focussed role of cleftee, however, does not

place any limitation on the syntactic role the NA may play in the lower clause of the cleft.

We find that the clefted non-canonical NA (as a kind of N’/NP) can play any role any other

N’ or NP can, included subject (15) and object (16), which are limited to N’ and NPs (pace

Fife 1993: 389).

(15) ys [tec a wr] a peris y wneuthur ynteu

is fair of man DIR caused its making it

‘It is a fair man that caused it to be made.’ W.SG.364

(16) ys [da a gedymdeith] a golleisti

is good of companion DIR you-lost

‘It is a good companion that you have lost.’ W.PKM.57.1

We also find them serving as predicates, which is open to both N’ (17) and AP (18).

(17) ys [ryued a antur] yw hwnn

is strange of adventure is this

‘It is a strange adventure that this is.’ W.SG.53

(18) ys [bychan a beth] uyd dy gywilyd di

is little of thing will-be your shame you

‘It is (a) little (thing) that your shame will be.’ W.PKM.79.1

Lastly, we find them playing a adverbial role (19), where primarily APs are possible. In

this position it alternates only with non-canonical APs. These results are summarized in

Table 1.

(19) [hir a beth] yd wyt yn triciaw

long of thing OBL you-are PT wait

‘(It is) a long thing that you are waiting (i.e. a long time).’ W.YBH.1848

As can be seen, while some uses of the non-canonical NA are incompatible with an AP

interpretation (as proposed by Fife) seemingly demanded by the structural dominance

relations of the internal syntax, others are precisely distributionally parallel to APs.

Orderly affect 427

Whatever the categorial status of this construction, it cannot be headed by A in all

distributions (15-17), though it could be in some (18-19). Therefore, strictly categorial

determination cannot produce a homogenous answer to the question; internal syntax does

not determine external syntax. There must be some other principle at work.

Table 1

Non-canonical NA NP N’ AP or A’

Subject cleft T T T –

Object cleft T T T –

Predicate cleft T – T T

Adverbial cleft T – – T

2.3. Subcategories: Referential and non-referential non-canonical NA

These distributional facts as well as paraphrase relationships lead us to believe that there

are two important classes of non-canonical NAs: Those with referential N heads which have

the distributions of and can be denotationally paraphrased by canonical NAs (15-17), and

those with non-referential N heads (the invariant peth ‘thing’ left untranslated in glosses)

which have the distributions of and can be denotationally paraphrased by As (18-19).

Therefore, this single construction (non-canonical NA) formally paradigmatically opposes

two types of ‘normal’ phrases, canonical N A (where the N head is referential) as well as

simple A (where it is not).

Referential (Nominal) Non-referential (Adjectival)

Canonical [NA N A] A

Non-canonical [NA A [PP a N]] [N

A A [PP a beth]]

The first class, the ‘referential’ non-canonical NA (in subject, object and predicate

roles), have canonical NA paraphrases which are more-or-less denotational synonyms of the

non-canonical NA : That is, formally ‘normalizing’ the syntactic relation between N and

A yields a normal NA which means more or less the same thing in denotational terms:

Non-canonical NAs with a referential N: Paraphrase relationships

Non-canonical NA = Canonical NA

tec a wr gwr tec ‘a fair man’

da a gedymdeith cedymdeith da ‘a good companion’

ryued a antur antur ryued ‘a strange adventure’

The second class, ‘non-referential’ non-canonical NPs (in predicate and adverbial roles),

cannot be so paraphrased. Instead, the N head peth ‘thing’ is left untranslated, and the non-

canonical NP is best paraphrased denotationally by a simple adjective or projection thereof

428 H. Paul Manning

(in predicate role) or an adverbial phrase (in the adverbial role) (see also Fife 1993: 380).

This latter function is one of the most common functions for a non-canonical NP, and

appears to be a general way to form adverbs from adjectives of expressive force (Evans

1976: 37n).

Non-canonical NAs with a non-referential N: Paraphrase relationships

Expressive NA = Adjective

bychan a beth bychan ‘(a) little’

hir a beth hir ‘(a) long (time)’

drwc a beth drwc ‘badly’

Table 2. Distributions of referential and non-referential non-canonical NA

Referential N Non-referential N

Subject T –

Object T –

Predicate T T

Adverbial – T

The environments where we find only ‘referential’ Ns are those where only NPs may serve:

Clefted subject and object of a finite verb (15-16, Table 2). We find non-referential and

referential Ns as a predicate of the copula (17-18): Here both phrases headed by N and

phrases headed by A may serve as the predicate. In fact, the two can be conjoined in this

function, and it seems that if there are conjoined adjectives modifying the same N head the

former in the series must make recourse to this non-referential strategy so that both can be

distinctively positioned (20).

(20) Non-referential N Referential N

[glew a beth] a [dewr a was] yw bown

valiant of thing and brave of boy is Bown

‘Bown is (a) valiant (thing) and a brave boy.’ W.YBH.466

In the adverbial environment there are only non-referential Ns (Table 2), although the

adverbial environment does not ban NPs in general.

If the referential non-canonical NA stands in a paraphrase relationship to a ‘normal’

NA, then the non-referential variant stands in a paraphrase relationship to a simple adjective

(or AP). The dummy head is provided as a way of allowing adjectives by themselves to

participate in this construction, so that the pragmatically focussed adjective can be formally

distinguished from the unmarked variety in terms of internal syntax. The adjective in both

referential and non-referential constructions acquires its distinctive form by virtue of its

position with respect to the noun head, whose referential status is irrelevant for these

purposes, it serves as the stuff from which paradigmatic form is fashioned. In the

Orderly affect 429

12 This observation is based on evidence drawn from my own biographical experiences ‘among the

linguists’; were this an anthropological journal I would cite my notes from the field.

referential construction, the formal expression of this adjectival category (paradigmatic

form) is parasitic on the expression of adjectival attribution (syntagmatic form). Since only

the pragmatic value of the adjective is at stake in this construction, the non-referential

construction provides a dummy ‘head’ (peth ‘thing’) with a vacuous adjectival modification

relation as the ‘matter’ to which this distinctive positioning of the adjective, as form, can

attach. Therefore, while the non-canonical NA with a non-referential N stands

distributionally as the expressive term in a semantic paraphrase relationship with a

denotational adjective (and not a denotational NA), in terms of its syntactic form it stands

in a paraphrase relationship with a ‘virtual’ canonical NA which can only be reconstructed

by analogy with referential non-canonical/canonical NA pairs.

This unusual construction (the ‘non-canonical’ N A) forms minimal pairs in

paradigmatic opposition to ‘canonical’ constructions (respectively NA and A) that can be

described formally in terms of internal and external syntax independently of the

grammatical categories that are encoded by it, as we have already seen. Such purely formal

minimal pairs contract ‘coding’ relationships with independently specifiable grammatical

categories of reference and modalized predication by virtue of participating in a

substantively organized “paradigm of configurations” (Silverstein 1993: 326) or a

“syntactic paradigm” (Matthews 1981: 267), organized substantively in the same way as

morphological paradigms are, but with configurational exponents:

Some paradigmatic structures of grammatical categories are expressed or coded in paradigms of

configurations, the constitution of which as paradigms...depends on having a formal-categorial

structure independent of, or definable without reference to, the grammatical categories crossed in

the paradigm. (Silverstein 1993: 326)

The canonical version of the NA codes nothing ‘extra’ by virtue of its syntax, it is

unmarked in this sense. As we will see below, such a form which is residual (unmarked)

from a grammatical categorial perspective also serves as the ‘default’, ‘elsewhere case’, as

a formal category (Zwicky 1994: 620-1). On the other hand, the non-canonical member of

the NA pair somehow “emphasizes” the adjective (Hemon 1975: 64; Fife 1993: 380). This

non-canonical NA construction alternates with and opposes the ‘unmarked’ canonical NA

as a ‘marked’ term of the opposition, by specifically and differentially signaling formally

the foregrounded ‘expressive’ pragmatic status of the adjective in grammatical categorial

terms. But what does ‘expressive’ mean?

4. Expressives: Form and content

The category of ‘expressive’ (and similar terms) in linguistic theory often seems like one

of those informal notions that we apply often not so much to explain a linguistic

phenomenon as to explain it away.12 Though there have been serious attempts to develop

a discussion of ‘expressives’ as a purely grammatical category, impressionistically, there

seems to be a widespread assumption amongst linguists that expressives are an inherently

430 H. Paul Manning

13 “The descriptive meaning of a judgement will be taken to refer to the speaker’s attitudes at the

time of speaking, even though the verb is not in the present tense. Thus ‘King John was bad’ becomes,

descriptively, not ‘I disapproved of King John,’ but rather, ‘I now disapprove of King John, who once

existed” (Stevenson 1944: 93). Michaelis and Lambrecht (1996: 239-40) note that the “deictic anchoring”

occurs on dimensions of “personal and temporal deixis”: “The notion of affective stance entails the presence

of someone making a judgement, and the speaker is the judge by default....Exclamatives in which the hearer

is the judge have a marked status” (ibid. 239), “With respect to temporal deixis, it appears that the affect

evoked by the main-clause predicator must hold at speech time, whether or not the state of affairs remarked

upon also holds at speech time.” (ibid. 240)

indescribable grammatical category that nearly always has unsystematic or unruly formal

expression. This (covertly or overtly) assumes that expressives belong to a class of (quasi-

)natural signs whose expression is essentially iconic (since the unruliness of the signifier

resembles that of the signified; cognition is coded by (digital) regularities, affect is coded

by (analogic) irregularities) (for recent restatements along these lines, see Danesh 1994;

Foolen 1997; Kryk Kastovsky 1997). Further confusion along these lines is added by the

tendency of some linguists to use the grammatical categorial label (‘expressive’) to name

a class of its formal exponents in some languages (e.g. Diffloth 1972, 1976). This

widespread informal assumption that expressive meaning will be coded by inherently

irregular or sporadic formal exponents (iconically) is strictly a priori; it has yet to be

demonstrated (Foolen 1997). What I will attempt to provide here will not be a full

pragmatic theory of expressive meaning (as, for example, Caffi and Janney 1994), but

rather simply a fairly clear metalinguistic vocabulary for just those grammatical categories

relevant to allow typological comparison of the MW constructions with those found in MB

and MC. Given a limited data set, some questions concerning the pragmatics of these

constructions are in principle unresolvable.

4.1. Expressives are indexicals

‘Expressives’ (here meaning all terms that have an evaluative component implying

potentiality for subjective affective engagement) carry an indexical ‘component’, in much

the same way as other indexicals (first explicitly noted, to my knowledge, by Stevenson

1944: 93; see also Caffi and Janney 1994: 346, 364-6; Oller and Wiltshire 1997: 39-41).

Michaelis and Lambrecht (1996: 239) use the phrase “deictic anchoring” to indicate that

expressives, involve, like other modalities of metapragmatic regimentation (e.g. indexical-

denotationals like deictics), stipulative (metapragmatic) reference to a hic et nunc indexical

origo (Bühler 1990 [1934]; Hanks 1992).13 It is this very indexical aspect that leads to their

being epistemologically opposed (like evidentials) as ‘private’ (indexing private affect

knowable to the speaker alone, an ‘inner’ deictic field (Caffi and Janney 1994: 364)) as

opposed to ‘public’ language (whose felicity can be ascertained by manifest public

conditions (Richards 1925: 131)). Most definitions of ‘expressive meaning’ therefore

assume that the very essence of this modality of indexicality is egocentric, speaker-centered

(e.g. Jakobson 1960: 354, Caffi and Janney 1994: 327). Partially this is so because the

expressive function posited by Jakobson (and earlier by Bühler (1990 [1934])) conflates a

function of language that foregrounds one component of the sign situation (speaker-

Orderly affect 431

centered) with a set of qualitative phenomena taken to be epistemically limited to speakers

(expression of subjective states of affect). However, since the indexical origo involved in

expressive meaning can shift to the addressee (Michaelis and Lambrecht 1996: 239), as

with deictics (Hanks 1992), we should separate out the modality of indexical expression

(expressiveness, subjective states of affect) from the indexical origo they are predicated of

(speaker, addressee, neither). Anticipating further results, I will use ‘emotive’ to denote

speaker-centered expressivity (at the token level), reserving ‘expressive’ to denote the more

general indexical potentiality of all these terms (at the type level). I will therefore use

‘evaluative’ and ‘expressive’ in roughly similar senses, bearing in mind that in my usage

evaluatives are a subset of expressives, which include all indexical terms potentially

susceptible to a pragmatic foregrounding of affective engagement, that is, both the

(positive/pejorative) evaluatives that have been of the most interest to philosophers, but

also intensives (augmentatives and diminutives, for example). Corresponding to these at

the type level are constructions which code token-level affectively engaging intensity (Caffi

and Janney 1994: 342; Fries 1995: 155), such as exclamative constructions (English Boy,

is it cold! ), where the intensity of the (sometimes purely denotational) quality of the

predication is affectively engaging.

4.2. Pragmatic categories: Emotive and ethical calibrations

Within social scientific approaches to language there has been recently an increasing

change of emphasis from the (speaker-centered) emotive and (addressee-centered)

evocative modalities of expressive meaning (Ogden and Richards 1923; Ayer 1946;

Stevenson 1944; Caffi and Janney 1994: 336), to more sociocentric ones (cf. Shweder

1984; Lutz 1988; Brenneis 1990; Irvine 1990). The transition from egocentric (‘solitary’,

subjective) to sociocentric (‘shared’, intersubjective) token-level implementation of an

intrinsically evaluative term like good seems to transform the modality of signalling pari

passu from “emotive” (“emotionally active”) to “descriptive” (“emotionally inactive” or

“ethical”) (Stevenson 1944: 83-4; see also Caffi and Janney 1994: 366). In order to clearly

delineate which indexical anchoring I intend, I will follow Stevenson and use the term

‘emotive’ to mean a very specific sort of (token-level) pragmatic implementation

(specifically egocentric here-now) of the intrinsic indexical potential of an inherently

evaluative term. Residually, any deviation from this reflexive hic et nunc pragmatic nexus

renders usage relatively ‘sociocentric’, for which I will use the term ‘ethical’. Sociocentric

expressive implementations of an evaluative term like good (glossable as ‘approved by

members of our community’), spatio-temporally (or socially) distantiated implementations

(‘approved in ancient Sparta’) or habitual implementations (‘usually approved by you and

me’)), are, for Stevenson, all equally ‘descriptive’ or ‘ethical’ (Stevenson 1944: 84, for a

parallel classification of indexicals see Voloshinov 1983: 12-3; Manning 2001). However,

the irreducible indexical potentiality of these expressive terms, whether used with an

egocentric ‘emotive’ calibration or a sociocentric ‘ethical’ calibration, is retained in some

form across these usages.

4.3. Metapragmatic categories: Thin and thick expressives

432 H. Paul Manning

14 Perhaps Stevenson means ‘acquires a potential for emotive meaning’, because here he would seem

to embrace a sociocentric definition of ‘emotive’ meaning for ‘thick’ ethical concepts which he specifically

rules out for ‘thin’ expressive concepts (which are ‘emotively active’ only when they are pragmatically

implemented in a certain way, and are otherwise ‘emotively inactive’ and therefore ‘descriptive’).

In spite of this ‘private’ aspect of terms with expressive potential (the indexical component

of potential for expressive meaning at the type level, or the egocentric emotive calibration

at the token level), these terms also have a relatively ‘public’ aspect to their meaning (the

denotational component at the type level, or sociocentric ethical implementations at the

token level) (Ortega y Gasset 1923; Mace 1934: 33-4; Stevenson 1944: 71ff; Black 1948:

116-7). The question is how to relate the indexical and the denotational components of

these terms at both the type and token level.

A very simple theory of expressive meaning, the Emotivist ‘Bah-Hurrah’ theory

attributed usually to Ayer (Ayer 1946: 108-9), paraphrases the expressive meaning of

utterances by periphrasis: “‘This is a bad painting’ is the same as ‘this painting - bah!’”

(Dickie 1971: 173). Among other things, this view collapses the distinction between

emotive and ethical pragmatic implementations of evaluatives in the favor of the former.

If we were to adapt this simple view to lexical semantics, we would assign the expressive

and denotational aspects to entirely distinct ‘components’ of meaning, so that positive

evaluatives might have appended to their denotational content (if there is any) an (indexical

expressive) feature (call it ‘[+Hurrah!]’) and negative evaluatives the opposed feature

(‘[+Bah!]’). We would then have a theory of hyponymy in lexical semantics in which, along

with so-called “cognitive hyponyms” (e.g. Cruse 1986: 291, the relation of dirty to mucky

(cf. Stevenson 1944: 82-3)), we would also have ‘expressive hyponyms,’ so that any

positively evaluative adjective is a hyponym of good (being a term whose ‘meaning’ is, for

the sake of argument, exhausted by its specification of [+Hurrah!])), and so on.

For a certain (rather small) segment of expressive vocabulary this may be more or

less a useful approach. These are so-called ‘thin’ ethical concepts (Williams 1985),

equivalent to Murdoch’s “primary and general moral words” (1970: 22), which do not

characterize a referent far beyond the expressive meaning, such as good. These are terms

“whose significance alludes especially and exclusively to the world of values: Good and

bad, better and worse....” (Ortega y Gasset 1923: 320). This is roughly the same as

Stevenson’s “independent” emotive meaning (1944: 72-3), which is “not a function of

descriptive meaning, but either persists without the latter or survives changes in it.” The

expressive potentiality of such terms follows from the fact that they specifically name or

stipulate their evaluative content as such.

We must also recognize a level of potentially expressive vocabulary which is so

because of the expressive nuances that follow from the denotation of the term. Here we

place so-called ‘thick’ ethical concepts (Williams 1985), or “normative-descriptive words”

or “specialized secondary value words (such as vulgar, spontaneous etc.)” (Murdoch 1970:

31). These are related to Stevenson’s ‘dependent’ emotive meaning (Stevenson 1944: 73),

which ‘is a function of descriptive meaning’ and which, for example “acquires a laudatory

emotive meaning partly because it refers, via its descriptive meaning, to something which

people favor” (Stevenson 1944: 71).14 The earliest equivalent distinction is found in the

work of (Ortega y Gasset 1923: 320-1). For him, words like noble, for example, “signify

both realities and values.” That is, on the one hand they have a clear referent (like non-

Orderly affect 433

evaluative adjectives such as red) as well as a positive evaluative value (unlike red). But

they do not have a positive value in general (as words like good do), but rather a

determinate positive value, one very distinct from that found in words like generous, useful,

elegant, this determinate positive value is dependent on the denotation, a kind of “halo of

meaning that surrounds the primary, realist, meaning of the word” (ibid.).

4.4. The ‘fade out’ of expressive transparency: A typology of expressive categories.

This lexical division between ‘thin’ and ‘thick’ expressives gives us some basis for

understanding qualitative changes in the basis or warrant of expressive meaning in the

expressive indexical vocabulary in a manner analogous to the qualitative changes of

metapragmatic value in organizing other indexical systems, as for example, differences of

metapragmatic transparency (Silverstein 1981a; Besnier 1993: 163) for indexicals

involved in reference (Silverstein 1981b), which organize what is usually informally called

the ‘animacy hierarchy’. Thus, if deictics, proper names, and even natural kind terms all

have an indexical component of some kind, the quality of that indexicality surely differs,

and this difference can be characterized in terms analogous to differences in “unavoidability

and transparency of metapragmatic reference” (Silverstein 1981b). Thus, informally,

deictics which refer to pragmatic roles created in-and-by the act of speaking ( I, You) are

more metapragmatically transparent than those deictics that are created by the discourse that

ensues (he, she, it); so proper names (which unavoidably index a baptismal moment of

speaking by which that name is conferred) are more metapragmatically transparent than

generic nouns; so too NPs that denote humans imply the potential incumbency to an event

of speaking, which NPs that denote non-humans do not, and so on. All such terms make

metapragmatic reference to the event of speaking, varying in the transparency of that

reference.

So too with expressive vocabulary, for inherently evaluative terms like good and

bad (or better still, Bah! and Hurrah! (cf. Besnier 1993: 163)) have a stipulative and

unavoidable potential for expressive use (although they may or may not be pragmatically

implemented with a specifically egocentric origo), and tend to be isolated in folk

metalinguistic (specifically metapragmatic) practice and linguistic ideology as the “primary

locus of affect in language” (Besnier 1993: 163). They therefore show a high degree of

“metapragmatic transparency”, which is “the degree of sameness between any

metapragmatic utterances that could be used to talk about a pragmatic form, and the

pragmatic form itself” (Silverstein 1981a: 11). Such transparency arises simply from the

fact that ‘thin’ evaluatives metapragmatically stipulate or name their own expressive

pragmatic value as such. As an illustration of a ‘folk metalinguistics’, Ayer’s ‘bah-hurrah’

theory of expressive meaning (above) relies precisely on the metapragmatic transparency

of ‘thin’ terms (bah! and hurrah!) metapragmatically to gloss the pragmatic value (positive

and negative evaluation) of ‘thick’ evaluatives with lower transparency (reducing the latter

to the former, just as he reduced ethical calibrations of such terms to emotive ones).

On the other hand, ‘thick’ evaluative terms like vulgar or genteel have an expressive

valuation and potential arising indirectly (less transparently) from denotational content that

may be rather more contingent on socio-historically and ideologically locatable factors

(hence “determinate positive or negative values” as opposed to “general” ones). Even so,

434 H. Paul Manning

the expressive valuation of such words is not completely independent from their

denotational content (cf. Black 1948: 116).

Even further removed from the core of expressive transparency would be those

words which have “emotional effects, whose sources cannot be discovered simply from a

knowledge of the definitions of the words themselves. Such words have emotional aurae,

which are by no means evident from their definitions” but which are due to their “cultural

history” (Henry 1936: 253-4).

A similar set of transitions exists for intensives. Intensive categories which refer

relatively transparently to the scale of intensity itself (big, great, little, small) are ‘thin’

intensives. These can be used in many languages as intensifiers for ‘thick’ intensives and

evaluatives (a little tired, great happiness). Essentially any scalar term will fall under this

latter category, since any reference to a scale thereby implies a subjective judgement and

the normative scale itself (the latter a relatively sociocentric form of indexical relation).

Therefore, denotational terms which lack inherent evaluation but are scalar can be deployed

in exclamative constructions which “express [emotive] judgements about situations that are

beyond the norm on some scale” (Fillmore 1999: 122). It is denotational abnormality with

respect to a normative scale that provides the affectively engaging indexical moment here.

Similarly, the token-level transition between egocentric, emotive, ‘solitary’

implementations of individual terms with an inherent expressive value to sociocentric,

ethical, ‘shared’ implementations is homologous to the type-level transition of potentiality

for such signaling between ‘thin’ (which reflexively and transparently stipulate their

expressive potential) and ‘thick’ lexemes (whose expressive potential arises indirectly and

contingently from their denotational content). The homology seems to be as follows:

Emotive usages are reflexive calibrations at the token level (pragmatically, as egocentric

token-reflexive calibrations), while ‘thin’ expressives are reflexive calibrations at the type

level (metapragmatically, in that they ‘name’ their own expressive content). ‘Thin’

expressives are more ‘unavoidably and transparently’ susceptible to ‘emotive’ token-level

calibrations.

Expressive meaning

Token-level (actual) Type-level (potential)

Calibration pragmatic metapragmatic

reflexive ‘emotive’ ‘thin’

non-reflexive ‘ethical’ ‘thick’

coding locus: grammatical lexical

Expressiveness at the type level (‘thin’ versus ‘thick’ inherently expressive lexemes (versus

non-expressive lexemes)) will correspond often to lexical selection in grammar, and

specifically to inherently expressive ‘thin’ expressive lexemes. On the other hand,

expressiveness at the token level (‘emotive’ versus ‘ethical’ uses of inherently expressive

lexemes) will correspond to grammatical alternations, open to any lexeme as long as it is

at least potentially expressive. Lastly, the locus of affect in language is not merely

evaluation ( positive versus pejorative) but also intensity, including oppositions like

Orderly affect 435

‘augmentative’ and ‘diminutive’ (Fries 1995: 155), but also intensity of purely denotational

scalar terms. The oppositions developed above apply equally to both of these qualitative

dimensions of affective coding.

For a coding view, grammatical categorial distinctions are of typological interest

(‘etically’) only insofar as they demonstrate their utility by being differentially coded

somewhere in some language (‘emically’) (Silverstein 1993; Manning 1995, 1997). We

will see that the MW construction signals emotive (exclamative) pragmatics at the token

level, and hence is usable with both ‘thin’ and ‘thick’ inherently expressive lexemes, while

a separate construction in MBC seems to code expressive potentiality at the type level, and

hence is limited to a lexically selected group of relatively ‘thin’ expressive lexemes.

5. Type-level and token-level expressives

5.1. Type-level expressives: Thin and thick evaluatives in MW

The MW non-canonical NP construction allows both ‘thin’ and ‘thick’ evaluative

adjectives. Naturally, the most common adjectives in this construction in token frequency

are the ‘thin’ expressives, dividable into two separate dimensions (Fries 1995: 155):

INTENSIVE (dividing into AUGMENTATIVE mawr ‘big, great’ and more marginally

DIMINUTIVE bychan ‘little’) (21), and EVALUATIVE (itself dividing into POSITIVE da

‘good’, and PEJORATIVE drwc ‘bad’) (22). These ‘thin’ expressive lexemes anchor two of

the dimensions of the category ‘expressive’ both in MW and arguably universally . With

these come the various constructionally stipulated ‘expressive hyponyms’ (labeling ‘thick’

expressive concepts) of these ‘thin’ expressives.

(21) INTENSIVE:

(21A) AUGMENTATIVE: mawr ‘big, great’, hir ‘long’, glut ‘diligent’, praf ‘thick,

stout, great’

(a) mawr a drueni (b) hir a beth

‘A great pity ‘ W.SG.258 ‘A long thing (i.e. a long time)’ W.YBH.1848

(c) glut a beth (d) praf a beth

‘A diligent thing (continually)’ W.PKM.7.16 ‘A great thing (i.e. a great deal)’ W.YBH.1106

(21B) DIMINUTIVE: bychan ‘small’.

(a) bychan a beth

‘A little (thing)’ W.PKM.79.1

(22) EVALUATIVE

(22A) POSITIVE: da ‘good’, tec ‘fair’, melys ‘sweet’, cadarn ‘strong’, iawn ‘true,

right’, glew ‘brave, stout’, dewr ‘brave’, gwybodus ‘well-informed’

(a) da varchawc (b) tec a wr

‘A good knight’ W.SG.294 ‘A fair man’ W.SG.364

436 H. Paul Manning

(c) melys a welioed (d) cadarn a ungwr

‘Sweet wounds’W.BD.166.1 ‘a strong man’ W.PKM.7.22

(e) glew a beth a dewr a was (f) iawn a beth

‘A stout (thing) and brave lad’ W.YBH.466 ‘(A) right (thing)’ W.PKM.8.11

(g) gwybodus a orderch

‘A well-informed sweetheart’ W.YCM.88.10

(22B) PEJORATIVE: drwc ‘bad’, tru(an) ‘poor, sad’, trist ‘sad’, anffydlawn

‘unfaithful’, anhebic ‘unexpected’, mileinyd ‘villainous’, meredic ‘foolish, stupid’,

cam ‘wrong’, dyhed ‘wretched’, ynuyt ‘foolish, mad’, ryued ‘strange’.

(a) drwc a gastell (b) tru a dynghetven

‘A bad castle’ W.SG.196 ‘A sad fate’ W.CA.6.136

(c) truan a dynghet (d) anffydlawn a wr

‘A sad fate’ W.CLlH.11.21 ‘An unfaithful man’ W.SG.407

(e) mileinyeid a swyd (f) meredic a wyr

‘A villainous post’ W.SG.424 ‘Foolish men’ W.KO.763

(g) a cham a beth (h) anhebic a gyflafan

‘And (a) false (thing)’ W.YBH.2461 ‘An unexpected crime’ W.PKM.43.22

(i) dyhed a beth (j) trist a galon

‘(A) wretched (thing)’ W.KO.133 ‘A sad heart’ W.YCM.96.12

(j) ynuyt a beth (k) ryued a antur

‘(A) foolish (thing)’ W.YBH.2570 ‘A strange adventure’ W.SG.53

The adjectives listed, while all inherently evaluative in some way, vary extraordinarily in

manner and degree, ranging from ‘thin’ evaluative adjectives (like da ‘good’, drwc ‘bad’)

to ‘thick’ evaluatives like gwybodus ‘well-informed’, ryued ‘strange’ or anhebic

‘unexpected’, which are rather less clearly intrinsically apt for expressive nuance.

Adjectives devoid of any potential expressive nuance are not attested. If the data is

representative, this implies that the construction differs from, for example, English

exclamatives of the Boy, was that cold! variety in not allowing purely denotational

adjectives.

5.2. Token-level expressives: Emotive pragmatics in MW

The ‘emotive’ pragmatics in which this kind of non-canonical NP in MW is involved tend

to be richly overdetermined by co-occurring exclamations, some of which foreground

intensity (for example, rof i a duw ‘between me and God’, dioer ‘faith’ ) (23a), others of

which focus on the evaluative dimension of affective involvement (for example och ‘alas’,

gwae vi ‘woe is me’) (23b).

Orderly affect 437

15 This makes it compatible with exclamative constructions (with mor) which foreground the

affectively engaging intensity of the quality denoted by the adjective (Michaelis and Lambrecht 1996;

Fillmore 1999: 122). Note that some exclamative constructions marking affectively engaging intensity can

co-occur with adjectives that have purely denotational values (Boy, was it red!), which is not the case with

this construction.

(23a) myn vympenn i heb y marchawc ys da chwedyl a dywedy di

‘By my head, said the knight, it is a good story that you tell.’ W.SG.410

(23b) och arglwyd heb y vorwyn ys mawr a bechawt awnaethost di amyui

‘Alas lord, said the maiden, it is a great sin that you did with me.’ W.SG.281

MW Non-canonical NPs are often found in construction with the particle mor ‘so’ when

it is used exclamatively (Evans 1976). The particle mor may intensify both evaluatives

(mauridic ‘magnificent (24b)) and intensives (mawr ‘great’ (24a)).

(24a) [mor vawr] [a ryfed]

so great of wonder

‘How great a wonder!’ W.T.79.18

(24b) [mor mauridic] [a kinran]

so magnificent of prince

‘How magnificent a prince!’ W.C.63.12

To conclude, the non-canonical construction operates on any adjective lexeme type that has

inherent expressive potential and stipulatively foregrounds that potential as an emotive

indexical calibration (foregrounding the speaker’s affective involvement in the evaluation

denoted by the adjective).15 This stands opposed to the lexically selected construction in

MB and MC which operates to code type-level expressive potentiality rather than token-

level expressive actuality, and hence tends to be limited to inherently expressive ‘thin’

expressive lexemes.

5.3. Comparison with MB and MC

In sharp contrast to MW, MC and MB have two cognate NA constructions, one of which

is lexical, where certain ‘thin’ evaluative (usually pejorative) adjectives must precede the

N (without the characteristic partitive structure involved in this construction), while the

other, operating on the residue of evaluative adjectives (essentially positive and intensive

evaluatives), is in free (semanticizable) alternation with the non-expressive ordering

(described above).

In Cornish, expressively pejorative (and some augmentative) adjectives obligatorily

precede the N, while other (positively) evaluative (and diminutive) adjectives generally

follow the residual unmarked ‘elsewhere’ pattern following the head N. There is variation:

Some adjectives, like mur ‘big, great’, tebel ‘evil, wicked’, ber ‘short’ can appear also in

postpositive position ( flehys mur ‘big children’ (P.168.3), an dewow tebel ‘evil gods’

438 H. Paul Manning

(O.1818), a termyn ver (R.706)). Certain other adjectives, such as bras ‘big, great’

(compare mur ‘great, big’) have only the post-positive order (gobar bras ‘great reward’

(R.672)). The allocation of adjectives to these patterns is ultimately lexical and

classificatory. The following examples give a particularly striking example of evaluative

antonyms showing opposite positionings (25).

(25) hager awel ha awel teg

ugly weather and weather fair

‘Ugly/bad weather and fair/good weather’ C.Pryce

(26) Pejorative adjective [A N] Non-pejorative antonym [N A]

drok venen flogh da

‘A bad woman’ C.O.221 ‘A good child’ C.O.664

an debel bobl branchys vas

‘Bad people’ C.O.1815 ‘Good branches’C.D.249

Augmentative Diminutive

mur varth flogh byhan

‘Great shame’ C.O.371 ‘A small child’ C.O.397

Two pairs of adjectival antonyms (Evaluative fals ‘false’ vs. guyr ‘true’, Intensive hir

‘long’ vs. ber ‘short’ (27)) deviate lexically from the norm (26), both being preposed.

(27) Pejorative adjective [A N] Non-pejorative antonym [A N]

fals marregion gwyr ures

‘False knights’C.R.607 ‘True judgement’ C.D.515

mar hir forth a ver termyn

‘So long a road’ C.O.734 ‘In a short time’ C . O . 1 3 6 2

Similarly in MB, we find that certain inherently expressive adjectives may precede the noun

in normal (non-partitive) constructions. Hemon (1975: 64) notes that such adjectives are

“monosyllabic, and express appreciation or depreciation, greatness, or smallness.” Hemon

cites meur ‘great’, berr ‘short’ , guir ‘true’, drouc ‘bad’, fals , fos, faux ‘false’, queah

‘poor’, goall ‘bad’. In other words, as in MC, they seem to center on the ‘thin’ expressives

of each pole of each expressive dimension posited above, especially pejorative concepts,

which form the core of the class. Some positive adjectives are occasionally attested

preposed: e.g. goar Mari ‘meek Mary’ (Jer.1444), clouar Mab Mari ‘sweet son of Mary’

(NG.1950). As in MC, some preposable adjectives are also attested in post-positive

position: Glau meur ‘big rain’ (Pm.67), an tut guir ‘the true people’ (BK.104), an den goar

‘the meek man’ (Pm.46), ma maestr clouar ‘my sweet master’ (J.82). The adjective coz,

kozh ‘old’ has the constructionally stipulated ‘thin’ pejorative sense of ‘bad, wretched’ in

prepositive position, and retains its denotational sense of ‘old’ in post-positive position,

thus indicating that minimal ‘expressive’/’denotational’ pairs existed for some adjectives,

sometimes within the same construct (28):

(28) ur c’hoz- varc’h coz

Orderly affect 439

the bad (lit. ‘old’) horse old

‘A wretched old horse’ B.SBI.2.120

Many of the same (cognate) adjectives occur in the same (cognate) construction in Cornish

(29), while Welsh does not seem to use this construction for any adjective but hen ‘old’.

(29) MB MC

meur truez mur lavur

‘Great compassion’ B.Nl.63 ‘Great labor’ C.O.299

berr termen a ver termyn

‘A short time’ B.J.18 ‘A short time’ C.O.1362

drouc scuezr drok venen

‘Bad example’B.M.548 ‘A bad woman’ C.O.221

hon guir Roue guyr gos

‘Our true king’ B.Nl.1 ‘True blood’ C.D.1506

fals marchador fals marregion

‘False merchant’ B.J.16 ‘False knights’ C.R.607

The allocation of adjectives to this construction in both languages appears to be primarily

lexical and arbitrary, since there are very few adjectives which have contrasting N A

constructions based on adjective position: The largely pejorative sense of the core of the

class is therefore essentially a lexical, classificatory, feature. By contrast, the ‘non-

canonical’ partitive construction in MB and MC is rather more ‘free’ to alternate with the

normal order (see above), as in MW, with the stipulation that only adjectives that do not

belong to the above lexical class may participate. In practice, this turns out to be

specifically positive (thin and thick) evaluatives in both MC and MB (see above for

examples (also Hemon 1975: 64-5)). In effect we have a neutralization of the ±Emotive

paradigmatic alternation (token-level pragmatics) for a lexically defined subclass of ‘thin’

potentially expressive terms (type-level pragmatics) (30).

(30) -Pejorative A +Pejorative A

-Emotive A [NA N A ] [N

A A N ]

+Emotive A [NA A [

PPa N]] --

The key difference between MB and MC, on the one hand, and MW, on the other, is that

the latter has generalized the possibility of systematically coding the ±Emotive opposition

to all potentially expressive adjectives (thin and thick, positive and pejorative), while

simultaneously eliminating for most adjectives the lexically defined preposed adjective NA

construction (which is found only with hen ‘old’ in Welsh, often in a pejorative or

diminutive sense). MW has simultaneously limited the external syntax of the construction

to appearing in cleft position. The grammatical and lexical categories posited for the

classification of expressive meaning therefore appear to have merit, in that they permit a

fairly straightforward characterization of the differences of distributions of cognate formal

categories in these languages.

440 H. Paul Manning

6. Conclusions: On the paradigmatic axis of syntactic relations

I have shown that P-Celtic languages in general have a formal syntactic distinction between

three realizations of NA, using this symbol to subcategorize the syntactic class N’, coding

‘adjectival modification of N’. The first opposition (found only in MC and MB) is a

(primarily) lexically mediated alternation selected by the adjective between preposed ([NA

A N]) and residual postposed ([ NA N A]) adjectives. While the former class has

associated with it (in MC and MB) primarily pejorative lexemes, the construction only

marginally forms minimal pairs with identical lexemes. The construction does not

necessarily foreground any token-reflexive emotive involvement, but simply classifies a

certain set of adjectives as having inherent expressive potential (‘thin’ evaluatives). In

addition, the residual class of N’s enters into a further opposition with a non-canonical NA

of ‘exocentric structure’ ([NA A [

PP a N]]). Adjectives are similarly opposed into a

canonical (simple A) and non-canonical NA ([NA A [

PP a N]]) construction with a dummy

head (peth ‘thing’), formally parasitic on the nominal construction. This opposition serves

to code the specific presence versus residual absence of emotive pragmatics. These

relations of lexical selection and expressive coding can be represented in a coding structure

(a structure which displays complex conditioning of dependent (formal) variables by

multivariate arrays of independent (substantive) variables (Manning 1995, 1997)) (42).

(42) Independent Variables Dependent variables

Type-level (‘thin’-‘thick’) Token-level (±Emotive)

Lexical selection (MCB) Expressive coding (ALL) NA form

[+ Pejorative] A --> [NA A N]

Elsewhere --> [+ Emotive] A --> [NA A [

PP a N]]

Elsewhere --> [NA N A]

The structure shows a binary paradigm, coding the pragmatic, token-reflexive grammatical

category [± Emotive] as applied to the adjective, ‘nestled’ within a lexical bifurcation

between the primarily pejorative ‘thin’ evaluative adjectives which as a type select preposed

order, and the remainder, which can undergo token-level differential expressive coding.

This latter level of coding applies only to MB and MC, in MW the coding of ±Emotive is

possible for all potentially evaluative adjectives (‘thin’ and ‘thick’, positive, intensive, and

pejorative). This coding structure shows graphically that the least marked NA type (in terms

of content and form) is the post-posed adjective (since this structure applies both to -

Emotive adjectives, and adjectives for which the opposition is irrelevant).

The ‘Coding View’ of this construction (1) denies that it is to be reduced to lexical

selection and (2) seeks to relate specificities of its formal behavior (both syntagmatically

and paradigmatically) to differentially encoded ‘meaning’. In this respect, we are then

viewing this as a ‘construction’ in the sense of Construction Grammar (see also Zwicky

1994):

Orderly affect 441

16 Virtually all earlier ‘holistic’ theories of grammar (Saussure’s, for example), as well as theories

of discourse (Bakhtin’s, for example), involve some sort of theory of internal relations (concepts like ‘value’,

‘dialogism’, respectively). Such a theory of internal relations has its antecedents in Post-Enlightenment

expressivist theories of language (Taylor 1985). When two items stand in some relation, but their identity or

nature is autonomous of and prior to that relation, they are externally related; when, however, their identity

or nature is in some way dependent on, or constituted by, that relation, they are internally related.

C is a CONSTRUCTION iffdef C is a form-meaning pair <Fi, Si> such that some aspect of Fi or

some aspect of Si is not strictly predictable from C’s component parts or from other previously

established constructions’ (Goldberg 1995: 4)

Whereas CG is almost exclusively concerned with univariate ‘form-meaning pairs’,

I am concerned here with multivariate ‘paradigms of configurations’. While CG does in

fact have a great deal to say about relations between constructions, it is less concerned with

systematic contrasts than with systematic partial similarities or partial inclusions (a notable

exception being Zwicky (1994)). In addition, while CG accounts talk in terms of relations

between form and ‘meaning’, these categories are seen to be in some sense basically

autonomous, contracting ‘external relations’ within the context of the construction that

pairs them. The coding view, by contrast, involves a theory of the mutual constitution of

form and meaning, a theory of ‘internal relations’, so that the specification of each involves

reference to the other.16

This perspective forces us in a principled manner to consider the non-canonical

structure alongside the canonical one, since without a specific difference in formal

categories there cannot be a specific and differential coding of grammatical categories (this

is precisely what the Saussurean notion of ‘value’ implies in a principled manner (Holdcroft

1991: 107ff)). CG is also explicitly indebted to aspects of the Saussurean notion of the

sign, but is primarily interested in Saussurean “relative motivation” (Goldberg 1995: 69),

which it captures primarily in terms of relations of “inheritance.” These paradigmatic

relations of partial similarity between constructions in absentia are precisely the mirror

image of paradigmatic relations of partial contrast foregrounded by the coding view. On

this level, CG seems to view the coding process as having, at least on its paradigmatic

dimension, the same systematicity as lexical coding: Each form-meaning pairing is an

absolutely arbitrary “listeme” (in the sense of Di Sciullo and Williams 1987), except to the

extent that it is relatively motivated by relations of inheritance (Goldberg 1995: 4-5;

Fillmore 1999: 115). The coding process is seen as an episodic series of “pairings” of

autonomous form and meaning (Zwicky 1994: 617), analogous to what Saussure criticized

as a ‘nomenclaturist’ view of language (Holdcroft 1991: 48). At best, there is an imperative

that differential form code differential content (Saussurean ‘value’) implied in the

“Principle of No Synonymy” (Goldberg (1995: 67); Zwicky (1994: 617) similarly stipulates

the equivalent of Saussurean ‘value’), but this does not specifically warrant the kinds of

local paradigmatic relations between syntactic constructions considered here.

CG is not unique in this regard. In most modern syntactic theories, the paradigmatic

relations of Saussurean grammar have languished in the “prison” (Di-Sciullo and Williams

1987) of the lexicon, while syntagmatic relations, to the extent that they are not ‘projected

from’ (or analytically ‘reduced to’) lexical particularities, only code pure compositional

semantic relations (Silverstein 1993: 328-9). Since the view of grammar espoused here is

specifically interested in the semiotic issue of the mutual constitution of categories of form

442 H. Paul Manning

and meaning (‘coding’), it seeks to avoid the circularity often implicit in such views (which

often reduce configurational coding relations to lexical ones by means of a lexical diacritic

(Silverstein 1993; Goldberg 1995)), and to reintroduce the paradigmatic dimension into

syntax.

CG, like the coding view espoused here, allows violations of ‘strictly categorial

determination’ of phrasal categories by the category of their lexical head. Interpreted rather

broadly, this implies that a given lexical head need not have a unique phrasal expansion (as

implied in theory, if not always in practice, in classic versions of X-bar theory). If this were

not so, then syntactic phrases can only have syntagmatic, compositional values and not

paradigmatic, contrastive values. If it is, it follows that there can be meaningful contrasts

between phrases (as shown here) just as there are between their lexical heads: What is not

predictable is meaningful. Syntactic paradigms are predicted by the latter assumption.

Unless, of course, one assigns the normal unmarked form of the phrase to the core grammar

represented by the X-bar theory, for example, and the abnormal marked phrase type to a

periphery based on ‘expressiveness’. The contrast becomes not a grammatical one, but a

contrast between essentialized components of orderly grammar and a disorderly periphery.

CG has also historically been interested in marginal constructions with equally

marginal pragmatics, including expressive constructions in syntax like the one here (see,

for example, Michaelis and Lambrecht 1996). In doing so its practitioners laudably seek

to redeem these marginal phenomena from the condescension of syntactic theories that

presume to have somehow located a ‘core’ of explainable syntactic phenomena amidst the

empirical phenomena of language (Culicover and Jackendoff 1999; Kay and Fillmore 1999;

Kay 2002). But CG identifies the ‘basicness’ of constructions not by reference to

paradigmatic relations of contrast, but syntagmatic interactions of inheritance (Michaelis

and Lambrecht 1996). I have argued, however, that such marginal phenomena are to be

analyzed against a backdrop of ‘unmarked’ phenomena with respect to which these

marginal constructions contract their ‘marked’ marginality. These internal relations

between constructions include not only positive ones like inheritance, but also negative

ones of differentiation captured with the traditional notion of paradigmatic contrast or

Saussurean value. If we do not invoke these paradigmatic relations between constructions

that link form and content not merely pairwise, but also crosswise in relations of

paradigmatic contrast, then we might as well entertain a theory of these marginal

constructions in which their abnormality of form is simply a function of their expressive

meaning. Under such a circumstance, syntactic and morphological features of these

marginal expressive constructions would deviate randomly from the phenomena of ‘core’

grammar, because they do not belong to the core grammar. If we do not treat marked

constructions (whatever they may code) in paradigmatic relation to their unmarked kin,

equally citizens of the sovereign state of grammar, we risk banishing them once again to

a marked hinterland of grammar. which can be studied without reference to the form of the

unmarked constructions, as a separate matter, a separate competence

The opposition between ‘core’ and ‘periphery’ in grammar (Culicover and

Jackendoff 1999; Kay and Fillmore 1999; Kay 2002) is homologous to the opposition

between ‘plain’ and ‘expressive’ morphology (or a fortiori, syntax) (Zwicky and Pullum

1987). There are equal and similar confusions in both, and both must be rejected by

linguistic theory. Rather than assigning meaningful formal variation to distinct unrelated

competences along such lines (always a possibility in a theory of external relations between

Orderly affect 443

constructions as form-meaning pairs, such as CG), we should attempt to see this variation

within the compass of a single competence, which requires the notion of a syntactic

paradigm (a theory of internal relations between constructions).

References

Ayer, A.J. (1946) Language, Truth and Logic. New York: Dover.

Baldi, Philip, and Chantal Dawar (2000) Creative processes. In G. Booij et al. (eds.),

Morphologie/Morphology 963-972.

Bat-El, Outi (2000) The grammaticality of ‘extra-grammatical’ morphology. In U. Doleschal & A. Thorton

(eds.), Extragrammatical and Marginal Morphology. Lincom Europa, pp. 61-81.

Besnier, N. (1993) Reported speech and affect on Nukulaelae atoll. In J. Hill & J. Irvine (eds.), Responsibility

and Evidence in Oral Discourse. Cambridge: Cambridge University Press, pp. 161-181.

Black, M. (1948) Some questions about emotive meaning. The Philosophical Review 57.2: 111-26.

Bloomfield, Leonard (1933) Language. New York: Holt, Rinehart and Winston.

Brenneis, Donald (1990) Shared and solitary sentiments: The discourse of friendship, play and anger in

Bhatgaon. In Catherine A. Lutz & Lila Abu-Lughod (eds.), Language and the Politics of Emotion .

Cambridge: Cambridge University Press, pp. 113-125.

Bühler, Karl (1990) (1934] Theory of Language: The Representational Function of Language (Foundations

of Semiotics 25). Amsterdam: John Benjamins Publishing Company.

Caffi, Claudia, and Richard Janney (1994) Toward a pragmatics of emotive communication. Journal of

Pragmatics 22: 325-373.

Cruse, D.A. (1986) Lexical Semantics. Cambridge: Cambridge University Press.

Culicover, Peter, and Ray Jackendoff (1999) The view from the periphery: The English comparative

correlative. Linguistic Inquiry 30.4: 543-571.

Danesh, Frantishek (1994) Involvement with language and in language. Journal of Pragmatics 22: 251-264.

Dickie, G. (1971) Aesthetics: An Introduction. London: Pegasus.

Diffloth, G. (1972) Notes on expressive meaning. Chicago Linguistic Society 8: 440-447.

Diffloth, G. (1976) Expressives in Semai. In P. Jenner et al. (eds.), Afroasiatic Studies, Part 1 (Oceanic

Linguistics Special Publications). Hawaii: The University Press of Hawaii, pp. 249-264.

DiSciullo A., and E. Williams (1987) On the Definition of Word. Cambridge, Mass: MIT Press.

Dowty, David (1996) Toward a minimalist theory of syntactic structures. In H. Bunt (ed.), Discontinuous

constituency. Berlin: Mouton de Gruyter, pp. 11-62.

Dressler, Wolfgang, and Lavinia Barbaresi (1994) Morphopragmatics: Diminutives and Intensifiers in Italian,

German and Other Languages. Berlin: Mouton de Gruyter.

444 H. Paul Manning

Elliot, Dale (1974) Toward a grammar of exclamations. Foundations of Language 11: 231-246.

Evans, D. Simon (1976) A Grammar of Middle Welsh. Dublin: The Dublin Institute for Advanced Studies.

Fife, James (1993) Decapitation in two Welsh adjectival constructions. Cognitive Linguistics 4.4: 371-394.

Fillmore, Charles J. (1999) Inversion and construction inheritance. In Gert Webelhuth et al. (eds.), Lexical

and Constructional Aspects of Linguistic Explanation. Stanford: CSLI, pp. 113-128.

Foolen, Ad. (1997) The expressive function of language: Toward a cognitive semantic approach. In Susanne

Niemeier & René Dirven (eds.), The Language of Emotions. Amsterdam: John Benjamins Publishing

Company, pp. 15-32.

Fries, Norbert (1995) Emotionen in der semantischen Form un der konzeptuellen Repräsentation. In Anadrás

Kertész (ed.), Sprache als Kognition - Sprache als Interaktion: Studien zum Grammatik-Pragmatik-

Verhältnis. Frankfurt: Peter Lang, pp. 139-181.

Goldberg, Adele (1995) Constructions: A Construction Grammar Approach to Argument Structure. Chicago:

University of Chicago Press.

Hanks, William H. (1992) The indexical ground of deictic reference. In A. Duranti & C. Goodwin (eds.),

Rethinking Context: Language as an Interactive Phenomenon. Cambridge: Cambridge University Press, pp.

43-76.

Hemon, Roparz (1975) A Historical Morphology and Syntax of Breton . Dublin: The Dublin Institute for

Advanced Studies.

Henry, J. (1936) The linguistic expression of emotion. American Anthropologist 38: 250-6.

Hill, Jane, and Judith Irvine (eds.) (1993) Responsibility and Evidence in Oral Discourse. Cambridge:

Cambridge University Press.

Holdcroft, David (1991) Saussure: Signs, Systems and Arbitrariness. Cambridge: Cambridge University

Press.

Humboldt, Wilhelm von (1988) On Language. Translated by Peter Heath. Cambridge: Cambridge University

Press.

Irvine, Judith (1982) Language and affect: Some cross-cultural issues. In H. Byrnes (ed.), Contemporary

perceptions of language: Interdisciplinary dimensions. Washington D.C.: Georgetown University Press, pp.

31-47.

Irvine, Judith (1990) Registering affect: Heteroglossia in the linguistic expression of emotion. In Catherine

A. Lutz & Lila Abu-Lughod (eds.), Language and the Politics of Emotion. Cambridge: Cambridge University

Press, pp. 126-161.

Jakobson, Roman (1960) Closing statement: Linguistics and poetics. In Thomas A. Sebeok (ed.), Style in

Language. Cambridge, Mass.: MIT Press, pp. 350-377.

Kay, Paul (2002) An informal sketch of a formal architecture for Construction Grammar. Grammars 5: 1-19.

Kay, Paul, and Charles Fillmore (1999) Grammatical constructions and linguistic generalizations: The What’s

x doing y? construction. Language 75.1: 1-33.

Orderly affect 445

Kryk-Kastovsky, Barbara (1997) Surprise, surprise: The iconicity-conventionality scale of emotions. In

Susanne Niemeier & René Dirven (eds.), The Language of Emotions . Amsterdam: John Benjamins

Publishing Company, pp. 155-169.

Lutz, Catherine A. (1988) Unnatural Emotions: Everyday Sentiments on a Micronesian Atoll and Their

Challenge to Western Theory . Chicago: University of Chicago Press.

Lutz Catherine A., and Lila Abu-Lughod (eds.) (1990) Language and the Politics of Emotion. Cambridge:

Cambridge University Press.

Mace, C.A. (1934) Representation and Expression. Analysis 1.3: 33-8.

Manning, H. Paul (1995). Fluid Intransitivity in Middle Welsh: Gradience, typology and ‘unaccusativity’.

Lingua 97: 171-194.

Manning, H. Paul (1996) What Welsh relatives are really like. In CLS 32: The Main Session. Chicago:

Chicago Linguistic Society, pp. 251-265.

Manning, H. Paul (1997) Indexical operators and scope relations in the Old Georgian relative clause.

Linguistics 35: 513-563.

Manning, H. Paul (2001) On social deixis. Anthropological Linguistics 43.3: 55-100.

Marouzeau, J. (1923) Langage affectif et langage intellectuel. Journal de Psychologie 20: 560-78.

Matthews, Peter H. (1981) Syntax. Cambridge: Cambridge University Press.

Maynard, Senko (1995) Interrogatives that seek no answers: Exploring the expressiveness of rhetorical

interrogatives in Japanese. Linguistics 33: 501-530.

Michaelis, L., and K. Lambrecht (1996) Toward a construction-based theory of language function: The case

of nominal extraposition. Language 72.2: 215-247.

Murdoch, Iris (1970) The Sovereignty of Good. Reading: Cox and Wyman.

Niemeier, Susanne, and René Dirven (eds.) (1997) The Language of Emotions. Amsterdam: John Benjamins

Publishing Company.

Norris, Edwin (1968) [1859] The Ancient Cornish Drama. (Two Volumes). New York: Benjamin Blom.

Ogden, C.K., and I.A. Richards (1923) The Meaning of Meaning. New York: Harcourt, Brace.

Oller, John W., and Anne Wiltshire (1997) Toward a semiotic theory of affect. In Susanne Niemeier & René

Dirven (eds.), The Language of Emotions. Amsterdam: John Benjamins Publishing Company, pp. 33-54.

Ortega y Gasset, J. (1923) Introducción a una estimativa. Revista de Occidente 4, October, 1923. Reprinted

in Obras Completas v. 6, 317-337. Madrid, Revista de Occidente.

Pryce (1972) [1790] Archeologia Cornu-Brittannica. Menston: Scolar Press Ltd.

Richards, I.A. (1925) Principles of Literary Criticism. New York: Harcourt Brace.

Richards, Melville (1940) Y frawdeg cymysg. [The mixed sentence]. Bulletin of the Board of Celtic

Studies.10.2: 105-116.

446 H. Paul Manning

Rosengren, Inger (1994) Expressive sentence types: A contradiction in terms. The case of exclamation.

Sprache und Pragmatik 33: 38-68.

Sapir, Edward (1921) Language. New York: Harcourt, Brace and World.

Sapir, Edward.(1949) [1933] Language. In D. Mandelbaum (ed.), Selected Writings of Edward Sapir in

Language, Culture and Personality. Berkeley: University of California Press, pp. 7-32.

Sapir, Edward (1994) The psychology of culture: A course of lectures reconstructed and edited by Judith T.

Irvine. New York : M. de Gruyter.

Shweder, R. (1984) Anthropology’s Romantic rebellion against the Enlightenment; or, there’s more to

thinking than reason and evidence. In R. Shweder & R. LeVine (eds.), Culture Theory: Essays on Mind, Self

and and Emotion. Cambridge: Cambridge University Press, pp. 27-66.

Silverstein, Michael (1981a) The limits of awareness. Sociolinguistics Working Papers 84. Austin, TX:

Southwest Educational Development Laboratory.

Silverstein, Michael (1981b) Case-marking and the nature of language. Australian Journal of Linguistics

1: 227-244.

Silverstein, Michael (1993) Of dynamos and doorbells: Semiotic modularity and the (over)determination of

cognitive representation. In CLS 29, Volume 2: The Parasession. Chicago: Chicago Linguistic Society, pp.

319-345.

Stankiewicz, E. (1964) Problems of emotive language. In T. Sebeok et al (eds.), Approaches to Semiotics.

The Hague: Mouton et co, pp. 239-276.

Stevenson, C.L. (1944) Ethics and Language. New York: Yale University Press.

Taylor, Charles (1975) Hegel. Cambridge.

Taylor, Charles (1985) Language and human nature. In Human Agency and Language. Cambridge, pp. 215-

247.

Taylor, Charles (1989) Sources of The Self. Harvard.

Voloshinov, V.N. (1973) [1929] Marxism and the Philosophy of Language. Cambridge, Mass.: Harvard

University Press.

Voloshinov, V.N. (1983) Discourse in life and disourse in poetry: Questions of sociological poetics. In Ann

Shukman (ed.), Bakhtin School Papers. Oxford: Oxford University Press, pp. 5-30.

Williams, B. (1985) Ethics and the Limits of Philosophy. Cambridge, Mass.: Harvard University Press.

Zwicky, Arnold (1987) Constructions in monostratal syntax. Chicago Linguistic Society 23: 389-401.

Zwicky, Arnold (1994) Dealing out meaning: Fundamentals of syntactic constructions. Berkeley Linguistics

Society 20: 611-625.

Zwicky, Arnold, and Geoffrey Pullum (1987) Plain morphology and expressive morphology. In J. Aske et

al. (eds.), Proceedings of the Thirteenth Annual Meeting of the Berkely Linguistic Society. Berkeley: Berkeley

Linguistic Society, pp. 330-340.